1. health and fitness
2. therapy

Clinical Psychology and Psychotherapy
Clin. Psychol. Psychother. 5, 126±144 (1998)
Comparative Efficacy of Treatments
for Post-traumatic Stress Disorder:
A Meta-Analysis
Michelle L. Van Etten1 and Steven Taylor2*
1
2
Department of Psychiatry, University of Michigan, USA
Department of Psychiatry, University of British Columbia, Vancouver, Canada
A meta-analysis was conducted on 61 treatment outcome trials for posttraumatic stress disorder (PTSD). Conditions included drug therapies
(TCAs, carbamazepine, MAOIs, SSRIs, and BDZs), psychological
therapies (behaviour therapy, Eye-Movement Desensitization and
Reprocessing
(EMDR),
relaxation
training,
hypnotherapy,
and dynamic therapy), and control conditions (pill placebo, wait-list
controls, supportive psychotherapies, and non-saccade EMDR control).
Psychological therapies had significantly lower drop-out rates than
pharmacotherapies (14% versus 32%), with attrition being uniformly
low across all psychological therapies. In terms of symptom reduction,
psychological therapies were more effective than drug therapies, and
both were more effective than controls. Among the drug therapies, the
SSRIs and carbamazepine had the greatest effect sizes, although the
latter was based upon a single trial. Among the psychological therapies,
behaviour therapy and EMDR were most effective, and generally
equally so. The most effective psychological therapies and drug
therapies were generally equally effective. Differences across treatment
conditions were generally evident across symptom domains, with little
matching of symptom domain to treatment type. However, SSRIs had
some advantage over psychological therapies in treating depression.
Follow-up results were not available for most treatments, but available
data indicates that treatment effects for behaviour therapy and EMDR
are maintained at 15-week follow-up. # 1998 John Wiley & Sons, Ltd.
INTRODUCTION
Post-traumatic stress disorder (PTSD) is characterized by three clusters of symptoms, which arise
after the person is exposed to a traumatic stressor.
The clusters are (1) recurrent reexperiencing of
the traumatic event (e.g. flashbacks, nightmares,
*Correspondence to: Steven Taylor, Department of Psychiatry,
2255 Westbrook Mall, University of British Columbia, Vancouver, B.C., Canada, V6T 2A1. Email: [email protected].
Contract grant sponsor: Medical Research Council of Canada.
Contract grant sponsor: National Institute on Drug Abuse.
CCC 1063±3995/98/030126±19$17.50
# 1998 John Wiley & Sons, Ltd.
intrusive thoughts) (2) avoidance of trauma-related
stimuli and numbing of general responsiveness,
and (3) persistent hyperarousal (e.g. hypervigilance,
exaggerated startle response: American Psychiatric
Association (APA), 1994). PTSD is often chronic,
and persists for at least 1 year after the trauma in
approximately 50% of cases (Davidson et al., 1996).
The most common precipitating events are combat
trauma, physical and sexual assault, natural disasters, and motor vehicle accidents (Breslau et al.,
1991; Davidson et al., 1991; Norris, 1992).
Community-based studies indicate that PTSD has
a lifetime prevalence of 1 to 14%, depending on
Meta-Analysis of PTSD Treatments
diagnostic methods and type of population, and not
surprisingly occurs at a much higher rate (3 to 58%)
in people who are at risk for exposure to traumatic
events (e.g. combat veterans, victims of natural
disasters or criminal violence: APA, 1994).
Several forms of treatment have been applied to
PTSD. Many treatments seem promising, although
the literature currently provides no clear indication
as to the method(s) of choice. Drug therapies used in
treating PTSD include tricyclic antidepressants
(TCAs), agents with anticonvulsant and moodstabilizing properties (e.g. carbamazepine), benzodiazepines (BDZs), monoamine oxidase inhibitors
(MAOIs), and serotonin specific reuptake inhibitors
(SSRIs). Drug therapies are based on the assumption
that exposure to trauma causes neurochemical
aberrations in mechanisms controlling arousal and
other aspects of emotional processing, and that
medications correct these aberrations. Changes in
the opioid, noradrenergic, dopaminergic, seratonergic, and hypothalamic±pituitary±adrenal axis
systems have all been implicated in PTSD (van der
Kolk, 1987; Friedman, 1991; Sutherland and Davidson, 1994). It is beyond the scope of the present article
to offer a more detailed discussion of the neural
structures, circuits, and neurotransmitters implicated in the various biochemical models of PTSD;
see Sutherland and Davidson (1994) for details.
Behavioural therapies (e.g. imaginal exposure)
and cognitive-behavioural therapies (e.g. stress
inoculation training: Veronen et al., 1978) for PTSD
were developed from conditioning and cognitive
theories. In an early formulation, Mowrer's (1960)
two-factor model was used to account for combatrelated PTSD (Keane et al., 1985). According to this
formulation, exposure to trauma produces a conditioned fear or anxiety response to trauma-related
stimuli. Escape and avoidance of trauma-related
stimuli are negatively reinforced (i.e. reinforced
because they provide short-term relief from distress),
and thereby prevent habituation from occurring.
Contemporary cognitive theories of PTSD are
consistent with neo-conditioning models (Rescorla,
1988), and emphasize expectations and appraisals
about the meaning of aversive experiences. Such
models include the emotional processing model
(Foa and Kozak, 1986; Foa et al., 1989) and similar
approaches (e.g. Chemtob et al., 1988; Litz and
Keane, 1989; Litz, 1992). These models propose that
PTSD symptoms arise from a fear structure stored
in long-term memory. The structure consists of a
network of information about stimuli, their meanings, and responses to those stimuli (e.g. autonomic
arousal, escape, avoidance). The traumatic experi# 1998 John Wiley & Sons, Ltd.
127
ence is thought to be so intense that it causes fearconditioning to a wide range of stimuli (e.g. sights,
sounds, odours, and bodily sensations associated
with the trauma). Such stimuli can serve as
reminders of the trauma (retrieval cues), thus
activating the fear structure and thereby producing
hyperarousal and intrusive recollections of the
trauma. Avoidance and numbing symptoms are
thought to arise from mechanisms for deactivating
the structure (Foa et al., 1992).
According to contemporary cognitive models,
PTSD can be reduced by exposing the person to
corrective information, which modifies the fear
structure. Behavioural and cognitive-behavioural
treatments are seen as effective means of producing
this change. An important ingredient in these
treatments is exposure to fear-evoking but objectively harmless stimuli. Some behavioral interventions also include cognitive restructuring, in which
the meaning of the trauma is examined. Training in
anxiety management skills is also provided (Foa
et al., 1989). Throughout this article we will use the
term `behaviour therapy' to include behavioral and
cognitive-behavioural treatments. We will examine
these treatments as a class of interventions rather
than evaluating specific types of treatment. This is
because there are insufficient trials to separately
examine each form of behavioural and cognitivebehavioural therapy. Our approach is similar to
other meta-analytic efforts to evaluate the efficacy
of broad classes of interventions (e.g. Lipsey and
Wilson, 1993).
Eye-movement desensitization and reprocessing
(EMDR) is a recent and controversial treatment that
entails imaginal exposure to traumatic images while
systematic saccadic eye movements are produced.
Saccades are typically induced by tracking a
therapist's finger as it is moved rapidly from side
to side (Shapiro, 1991). Coping statements also are
introduced while the scene is being imagined.
Treatment typically takes one to four sessions.
Recently, Shapiro (1995) suggested that eye movements in EMDR can be replaced with other lateral,
stereotypic, motor movements. Shapiro (1995)
postulated that exposure to trauma produces
neuronal changes and disruption of a physiological
balance between excitatory and inhibitory systems
in the brain, which prevents appropriate processing
of traumatic memories. EMDR purportedly restores
this balance and reverses the neural pathology, and
in so doing, allows appropriate reprocessing and
integration of the traumatic memories (Shapiro,
1991, 1995). The theoretical underpinnings of
EMDR have been criticized by several writers
Clin. Psychol. Psychother. 5, 126±144 (1998)
128
(e.g. Lilienfeld, 1996; McNally, 1996). It may be that
EMDR is an effective treatment, but not for the
reasons proposed by Shapiro (1995).
Another therapy used for PTSD includes relaxation training, which is aimed at reducing hyperarousal (e.g. Vaughan et al., 1994). Other treatments
include hypnotherapy and psychodynamic psychotherapy, which are aimed at uncovering and
resolving unconscious conflicts arising from the
traumatic events (e.g. Brom et al., 1989).
Despite the many treatments used for PTSD,
none have been established as treatments of choice,
and there has been only one previous attempt to
quantify the relative efficacy of these interventions.
Otto et al. (1996) reported an effect-size analysis for
14 treatment-outcome studies representing 20 trials
of drug therapy or psychological therapy. Studies
were published or presented between 1991 and 1994.
Only randomized controlled trials were included. On
measures of PTSD symptoms, general anxiety, and
depression, the drug therapies with the largest effect
sizes were fluoxetine and amitriptyline. Behavioural
therapies tended to have larger effect sizes than these
drug therapies, and were associated with less
attrition. Follow-up data were not examined, and
no trials of EMDR were included.
These findings were based on a small number of
trials, and so the results should be regarded with
caution. Moreover, there are several major methodological concerns with Otto et al.'s (1996) study. They
computed each effect size by subtracting the mean
of the posttreatment treatment group from the mean
of the posttreatment control group, and then
dividing by the standard deviation of the control
group. The problem with this approach is that it
ignores pretreatment differences between treatment
and control groups. The trials typically consisted of
small numbers of participants (e.g. Ns of 8 to 16).
With such small samples it is likely that random
assignment of participants to treatment versus
control groups would often fail to equate groups
on pretreatment severity. This means that some of
the effect sizes may actually represent pretreatment
differences rather than differences in the efficacy of
treatment and control conditions. Moreover, Otto
et al. compared treatments against different types of
controls. Drug therapies were compared to pill
placebo whereas psychological therapies were typically compared to waiting-list controls. Thus, the
comparison of treatments was confounded by the
use of different types of controls.
A further concern with the Otto et al. (1996)
study is that they computed effect sizes across
scales, thus combining data from self-report and
# 1998 John Wiley & Sons, Ltd.
M. L. Van Etten and S. Taylor
observer-rated scales. Observer-rated scales typically yield larger effect sizes than self-report scales
(e.g. Lambert et al., 1986; Taylor, 1995). If observerrated scales were more likely to be used in studies
of some treatments (e.g. drug therapies) than in
others (e.g. behaviour therapy), then the comparison between treatments will be confounded by
differences in assessment method.
Otto et al. (1996) did not include uncontrolled
trials, and thereby excluded many studies from their
analysis. For a given type of treatment (e.g.
behaviour therapy), the effect sizes of these trials
can be compared with those of controlled trials in
order to determine the comparability of controlled
and uncontrolled trials. If the mean effect sizes for
controlled and uncontrolled trials do not differ, then
uncontrolled trials can be included, thereby increasing statistical power (Hunter and Schmidt, 1990).
The purpose of the present study was to further
investigate the comparative efficacy of PTSD treatments, using a broader range of treatments than
those examined by Otto et al. (1996). We also
intended to circumvent the methodological concerns inherent in the latter study. We used metaanalysis to empirically evaluate the relative efficacy
of treatments for PTSD. Our aims were (1) to identify which classes of treatment are more effective
than wait-list controls and placebo; (2) to determine
whether some classes of treatment are more effective
than others; and (3) to determine whether treatment
gains are maintained at follow-up.
METHOD
Inclusion and Exclusion Criteria
English-language articles published, unpublished,
or presented at conferences from 1984 to 1996 were
located from Medline, the PILOTS Database,
Psychological Abstracts, Current Contents, conference programs, recent journal issues, and secondary
sources (e.g. narrative reviews, book chapters), and
by contacting PTSD researchers. Articles were
included if the following criteria were met: (1) all
participants were diagnosed with PTSD according
to DSM III, DSM III-R, or DSM-IV criteria, as
assessed by structured or unstructured clinical interviews. (2) Five or more participants were included in
each trial. (3) Sufficient information was provided to
compute effect sizes (or necessary additional data
was supplied by the authors). (4) Outcome was
presented in terms of self-report or observer-rated
measures for one or more of the following variables:
Clin. Psychol. Psychother. 5, 126±144 (1998)
Meta-Analysis of PTSD Treatments
intrusions, avoidance, total PTSD severity, depression, and anxiety. These variables were selected
because they are the ones most commonly used
to assess outcome in treatments of PTSD. (5) The
outcome measures had acceptable levels of
reliability and validity, as reported in the outcome
study or in previous reports.
A total of 41 studies were located, yielding 68
treatment-outcome trials. Three trials from three
different studies were of various inpatient treatments that were sufficiently heterogeneous and/or
poorly described so as to prevent interpretation
of the data as a distinct treatment class. These
were therefore excluded from the analysis, leaving
65 trials. In 61 of the these trials, most participants
had chronic PTSD, and four trials included only
people with acute PTSD. Chronic PTSD is defined
as duration of symptoms of 3 months or longer
(APA, 1994). The four trials based on acute PTSD
included two behaviour therapy trials, an assessment-only trial, and a trial of relaxation training.
The mean duration between the trauma and
commencement of each outcome trial ranged from
3 to 12 weeks in the studies of acute PTSD,
compared to approximately 6 years in studies of
chronic PTSD. Compared to the effect sizes for
chronic PTSD, the effect sizes for the four trials of
acute PTSD were statistical outliers. The large effect
sizes for acute PTSD may reflect spontaneous
remission, which is more common in acute than
chronic PTSD (Foa, 1994; Rothbaum et al., 1992).
Thus, our meta-analysis consisted of 61 trials from
39 studies of chronic PTSD. These are listed in
Table 1. All treatments were provided in individual
format with the exception of one behaviour therapy
trial, which used a combination of group and
individual treatment (Frueh et al., 1996). An
appendix listing the studies that were excluded,
and reasons for exclusion, is available on request.
Of the 61 trials included in the meta-analysis,
36 were from studies in which two or more conditions (e.g. TCA versus placebo) were compared.
Five studies used crossover designs, where participants completed one condition followed by a
waiting period, and then completed another condition (Shestatzky et al., 1988; Reist et al., 1989;
Braun et al., 1990; Pitman et al., 1996a; Rothbaum
et al., 1996). To avoid the problem of confounding
within- and between-subject variance, and to avoid
possible problems of carry-over effects from one
treatment to another, we included only the first
active treatment trial (i.e. drug or psychological
treatment) from each crossover study. To illustrate,
Shestatzky et al. (1988) used a cross-over design
# 1998 John Wiley & Sons, Ltd.
129
with one condition consisting of phenelzine
followed by pill placebo. Here, we included only
the phenelzine condition.
Of the trials included in the meta-analysis, six
were TCA treatments, which included desipramine
(n (number of trials) ˆ 2; mean dose (i.e. mean
dose at the end of treatment) ˆ 200 mg/day),
imipramine (n ˆ 2; mean dose ˆ 242 mg/day),
amitriptyline (n ˆ 1; mean dose ˆ 175 mg/day),
and trazadone (n ˆ 1; mean dose ˆ 300 mg/day).
Although carbamazepine is structurally similar to
TCAs, we classified it separately because it appears
to have different pharmacologic properties to
conventional TCAs. In addition to its anti-seizure
effects, it is thought to reduce problems of impulse
control (Coccaro and Siever, 1995), which in turn
raises the question of whether it plays an important
role in the reduction of unwanted, intrusive,
trauma-related thoughts (Lipper, 1990). One trial
of carbamazepine that was suitable for inclusion
was located (mean dose ˆ 661 mg/day).
Seven MAOI treatments were included, consisting
of phenelzine (n ˆ 6; mean dose ˆ 60 mg/day) and
brofaromine (n ˆ 1; mean dose ˆ 150 mg/day). BDZ
treatment consisted of a single trial of alprazolam
(mean dose ˆ 3.75 mg/day). Four SSRIs trials
included fluoxetine (n ˆ 2; mean dose ˆ 60 mg/
day), fluvoxamine (n ˆ 1; mean dose ˆ 150 mg/
day), and sertraline (n ˆ 1; mean dose ˆ 105 mg/
day). All patients in all drug trials were on
medication when assessed at posttreatment.
Thirteen behavioural therapy trials were included.
They generally entailed some type of exposure
therapy (n ˆ 11), with some of these using imaginal
exposure (n ˆ 4) and others using both imaginal and
in-vivo exposure (n ˆ 7). Some behavioural therapies
also included stress-inoculation training (SIT: n ˆ 3).
As mentioned earlier, we examined behavioural
therapies as a group (which included cognitivebehavioural treatments), rather than separately
examining each `type' of behavioural intervention.
This was because there were insufficient trials to
conduct a more fine-grained analysis. Thus, our
meta-analysis was directed toward examining behaviour therapy as a class of interventions, which is
similar to the way in which other meta-analyses
have examined classes or groups of interventions
(see, for example, Lipsey and Wilson's (1993) metaanalysis of very broad classes of psychological and
educational interventions).
EMDR therapies were also examined as a class of
therapies, consisting of 11 trials. Although EMDR
has been modified since it was first described by
Shapiro (1989), the initially proposed elements of
Clin. Psychol. Psychother. 5, 126±144 (1998)
130
# 1998 John Wiley & Sons, Ltd.
Table 1. Trials included in meta-analysis
Author(s)
Condition
N completers
% dropout
Trial duration
(weeks)
Baker et al. (1995)
Self-report measures
Observer-rated measures
Brofaromine 150 mg
Pill placebo
56
58
5
2
12
12
DTS, IES,
DTS, IES
CAPS
CAPS
Braun et al. (1990)
Alprazolam 3.75 mg
10
38
5
IES
15
14
19
16
IES,
IES,
IES,
IES,
HAM-A, HAM-D,
PTSD Scale
Ð
Ð
Ð
Ð
Brom et al. (1989)* {
Exposure therapy
Hypnotherapy
Dynamic psychotherapy
Waiting-list control
29
29
31
23
11
11
11
11
STAI-T
STAI-T
STAI-T
STAI-T
Burstein (1984)
Imipramine 260 mg
10
33
3
IES
Ð
Carlson et al. (in press)* {
EMDR
Relaxation training
Supportive psychotherapy
10
13
12
0
8
0
12
12
6
BDI, IES, MISS, STAI-T
BDI, IES, MISS, STAI-T
BDI, IES, MISS, STAI-T
CAPS
CAPS
CAPS
Cooper and Clum (1989)
Exposure therapy
Supportive psychotherapy
8
8
27
27
10
10
BDI, STAI-T
BDI, STAI-T
Ð
Ð
Phenelzine 52.5 mg
7
36
6
IES
HAM-A
Davidson et al. (1990)
Amitriptyline 175 mg
Pill placebo
17
16
29
29
8
8
IES
IES
HAM-A, HAM-D, SI-PTSD
HAM-A, HAM-D, SI-PTSD
Devilly and Spence (1996)
EMDR
Supportive psychotherapy
No-saccade control
19
16
16
Ð
Ð
Ð
2
2
2
BDI, MISS, STAI-T
BDI, MISS, STAI-T
BDI, MISS, STAI-T
Ð
Ð
Ð
Foa et al. (1996)* {
SIT
Exposure therapy
SIT ‡ exposure therapy
Waiting-list control
19
22
22
15
Ð
Ð
Ð
Ð
9
9
9
9
BDI,
BDI,
BDI,
BDI,
STAI-T
STAI-T
STAI-T
STAI-T
PSS-I
PSS-I
PSS-I
PSS-I
Foa et al. (1991)* {
Exposure therapy
SIT
Supportive psychotherapy
Waiting-list control
10
14
11
10
29
18
21
0
5
5
5
5
BDI,
BDI,
BDI,
BDI,
RAST,
RAST,
RAST,
RAST,
PSS-I
PSS-I
PSS-I
PSS-I
Forbes et al. (1994)* {
EMDR
BDI, IES
Frueh et al. (1996)
Trauma mgt therapy
Hertzberg et al. (1996)
Hickling and Blanchard
(1997)*
8
Ð
4
15
27
17
Ð
CAPS, HAM-A
Trazadone 150 mg
6
0
16
DTS
CAPS
Exposure therapy
8
17
10
Ð
CAPS
Ð
SI-PTSD
BDI
BDI, STAI-T
BDI, STAI-T
HAM-A, HAM-D
PTSD Symptom Checklist
PTSD Symptom Checklist
Jensen (1994){
EMDR
13
14
3
Kauffman et al. (1987)
Keane et al. (1989)
Desipramine 200 mg
Flooding
Waiting-list control
8
11
13
Ð
0
0
4
14
20
Kosten et al. (1991)
STAI-T
STAI-T
STAI-T
STAI-T
SI-PTSD
Imipramine 225 mg
23
48
6
Covi Anx, IES, Raskin Dep
HAM-D
Phenelzine 68 mg
Pill placebo
19
18
48
48
6
6
Covi Anx, IES, Raskin Dep
Covi Anx, IES, Raskin Dep
HAM-D
HAM-D
M. L. Van Etten and S. Taylor
Clin. Psychol. Psychother. 5, 126±144 (1998)
Davidson (1987)
EMDR
8
12
1
Lerer et al. (1987)
Phenelzine 60 mg
Phenelzine 60 mg
Phenelzine 60 mg
6
7
9
12
36
64
6
11
16
Lipper (1990)
Carbamazepine 661 mg
10
9
5
BDI, IES, IPAT, PTSD Index HAM-A, HAM-D, PTSD Sx
cklst
Marcus et al. (1996)* {
EMDR
67
1
3
BDI, IES, M-PTSD
Marmar et al. (1996)
Fluvoxamine 150 mg
10
9
10
IES, SCL Anx & Dep, SRRS
Ð
Mollica et al. (1990)
Supportive psychotherapy
26
34
24
Ð
Hopkins Anx & Dep Scales
Montgomery and Ayllon
(1994)
EMDR
6
Ð
2
BDI
Ð
IES
CAPS, HAM-A, HAM-D
BDI, IES
CAPS
Ð
Ð
Ð
HAM-A, HAM-D, PTSD Scale
HAM-A, HAM-D, PTSD Scale
HAM-A, HAM-D, PTSD Scale
Ð
Nagy et al. (1993)
Fluoxetine 80 mg
10
63
10
Pitman et al. (1996a)* {
Pitman et al. (1996b)* {
Flooding
20
0
6
IES, MISS, PTSD Sev Scale
Ð
EMDR
17
30
6
IES, MISS
CAPS
HAM-A, HAM-D
Reist et al. (1989)
Desipramine 200 mg
18
22
4
BDI, IES
Richards et al. (1994)
Rothbaum (in press)* {
Exposure therapy
13
7
8
BDI, IES, PTSD Checklist
Ð
EMDR
Waiting-list control
10
8
14
14
3
3
BDI, IES, RAST, STAI-T
BDI, IES, RAST, STAI-T
PSS-I
PSS-I
Rothbaum et al. (1996)
Scheck et al. (in press)* {
Sertraline 105 mg
5
Ð
12
IES
CAPS
EMDR
26
30
2
BDI, IES, PENN, STAI-T
Ð
Shestatzky et al. (1988)
Phenelzine 60 mg
10
54
4
IES
PTSD Scale, HAM-A, HAM-D
Vaughan and Tarrier (1992)
Wilson et al. (1995, 1997)* {
Image habituation
10
Ð
10
BDI, IES
Ð
EMDR
32
Ð
3
IES, STAI-T
Ð
Fluoxetine 40 mg
Pill placebo
21
27
36
13
5
5
Ð
Ð
CAPS
CAPS
van der Kolk et al. (1994)
Meta-Analysis of PTSD Treatments
# 1998 John Wiley & Sons, Ltd.
Lazrove et al. (1996)* {
131
Clin. Psychol. Psychother. 5, 126±144 (1998)
Doses for drug trials refer to the mean dose attained by the end of treatment. `Ð' refers to data not reported (for % dropouts) or not used (for measures).
* Psychotherapy trial reporting treatment fidelity check.
{Psychotherapy trial reporting level of therapist training.
BDI, Beck Depression Inventory; CAPS, Clinician Administered PTSD Scale; Covi Anx, Covi Anxiety Inventory; DTS, Davidson Self-Rating Trauma Scale; EMDR, Eye Movement
Desensitization and Reprocessing; HAM-A, Hamilton Anxiety Scale; HAM-D, Hamilton Depression Scale; IES, Impact of Event Scale; IPAT, IPAT Anxiety Scale; MISS, Mississippi
Scale for PTSD; M-PTSD, Modified PTSD Scale; PE, Prolonged Exposure; PENN, Penn Inventory for PTSD; PSS-I, PTSD Symptom ScaleÐInterview Form; Raskin, Raskin Depression
Scale; RAST, Rape Aftermath Symptom Test; SCL Anx & Dep, Anxiety and depression scales from the SCL-90-R; SIT, Stress Innoculation Training; SI-PTSD, Structured Interview for
PTSD; SRRS, Stress Response Rating Scale; STAI-T, Trait version of State-Trait Anxiety Inventory.
132
treatment have remained unchanged: i.e. imaginal
exposure with concomitant lateralized movements, along with coping statements (Shapiro,
1995; personal communication, 1 November 1996).
Relaxation therapy included biofeedback-guided
relaxation (n ˆ 1). Other treatment trials included
hypnotherapy (n ˆ 1) and psychodynamic therapy
(n ˆ 1).
Control groups included pill placebo (n ˆ 4),
wait-list control (n ˆ 5), a non-saccade control for
EMDR studies (n ˆ 1), and supportive psychotherapy (n ˆ 5). The non-saccade condition is an
EMDR control in that it includes no eye saccades
nor any other oscillating stimulation. The supportive psychotherapy condition included three trials
where participants received standard supportive
therapy at a VA medical centre, one trial described
as a supportive control in which subjects received
general therapist support and some teaching in
general problem-solving, and one trial in which
participants met weekly with a social support
service team. Four of the five supportive psychotherapy trials were described by their authors as
control conditions. These trials can therefore be
regarded as attention placebos.
Statistical Procedures
Effect sizes were calculated according to Cohen's
(1988) d statistic. For each trial the magnitude of
change from pre- to posttreatment was defined
as …Mpre ÿ Mpost †/SDpooled , where SDpooled ˆ
p
‰…SD2pre ‡ SD2post †=2Š. The magnitude of change
from pre-treatment to follow-up was defined by
replacing Mpost with Mfollow-up , and SDpost with
SDfollow-up . For the outcome measures used in
the present study, positive effect sizes represent
improvements in PTSD and other symptoms (i.e.
reductions in problem severity), whereas negative
effect sizes indicate a worsening of symptoms.
Effect sizes were based on completer analyses rather
than end-point or intent-to-treat analyses. In other
words, effect sizes were based on pre- and posttreatment data for participants completing each
trial. This was because most trials only reported
data for treatment completers.
There are a number of different formulae for
computing effect sizes (for examples, see Smith et al.,
1980; Wolf, 1986) and none has been established as
a gold standard. We selected the above-mentioned
effect size because the same or very similar effect
size formulae are commonly used (e.g. Christensen
et al., 1987; Taylor, 1996; Abramowitz, 1997a,b)
and because it provides an effect size for each trial,
# 1998 John Wiley & Sons, Ltd.
M. L. Van Etten and S. Taylor
rather than an effect size defined as the posttreatment difference between a treatment and control
trial. Thus, we were able to include uncontrolled
studies in the meta-analysis, thereby increasing the
number of trials and statistical power to detect
differences between treatments. (Note also that we
were able to determine whether the effect sizes of
controlled studies differed from those of uncontrolled studies). The different formulae for effect
sizes can differ in the magnitudes of obtained
effects, and the effect size used in the present study
tends to be larger than effect sizes computed
according to other methods (Abramowitz, 1997a).
Accordingly, one should not interpret in isolation
an effect size for a given treatment. The meaning of
the effect size is determined by comparing it to the
effect sizes for other treatments.
Regardless of the method of computing effect
sizes, a further concern that we will examine is the
possibility of inflated effect sizes due to the `file
drawer' problem (Rosenthal, 1979). We use the
procedures outlined by Hunter and Schmidt (1990,
pp. 512±513) to determine whether this was a
problem for the treatments examined in the present
study. The file drawer problem is a publication bias
in which studies obtaining significant findings and
large effect sizes are published, whereas findings
obtaining null results are unpublished. To determine the likelihood of this bias, the fail-safe N is
computed (Orwin, 1983), which is the number of
unpublished trials obtaining zero effect sizes that
are required to reduce an obtained mean effect size
to a trivial level.
If a large number of unpublished trials are
required, then it is unlikely that the obtained effect
is biased by the file drawer problem. The number of
unpublished or unobtained trials obtaining zero
effect sizes is defined by k[(dk/dc) 7 1], where
k ˆ the number of obtained trials, dk ˆ mean obtained effect size, dc ˆ the trivial value to which the
obtained effect would be reduced. Note that dc
cannot equal 0, because dk/dc would be undefined.
Orwin (1983) suggested that a small effect size
(i.e. 0.200) would qualify as a trivial value. However, such an effect size is considered non-trivial by
some meta-analysts (e.g. Lipsey and Wilson, 1993).
Accordingly, we defined a trivial effect size as
0.050.
Assessment
Previous
meta-analyses
have
shown
that
interviewer-rated scales consistently yield larger
effect-sizes than self-report scales (e.g. Lambert
Clin. Psychol. Psychother. 5, 126±144 (1998)
Meta-Analysis of PTSD Treatments
et al., 1986; Taylor, 1995). This may reflect the
greater sensitivity of interviewer-rated scales, or it
may be an artifact reflecting interviewer bias (i.e.
the interviewer typically knows whether or not the
assessment is a pre- or posttreatment evaluation,
and therefore may be biased by expecting comparatively lower symptom scores at the posttreatment
assessment). Systematic bias in computing effect
sizes can occur if some types of treatments
are evaluated with interviewer-rated measures
(e.g. drug therapies), while others are evaluated
with self-report measures (e.g. psychological therapies). Thus, we calculated treatment effect sizes
separately for interviewer-rated and self-rated outcome variables (see Table 2).
The outcome measures used in each study are
presented in Table 1. As the table shows, scores for
self-reported intrusions, avoidance and total PTSD
symptoms were obtained from the Impact of Event
Scale (Horowitz et al., 1979), the Mississippi Scale
for Combat-related PTSD (Keane et al., 1988), and
various DSM-tailored measures such as the PTSD
Symptom Checklist (Richards et al., 1994), PTSD
Index (Lipper, 1990), and Modified PTSD Scale
(Saunders et al., 1990). Self-reported depression was
typically assessed by the Beck Depression Inventory
(Beck et al., 1979), and self-reported anxiety was
typically assessed by the State-Trait Anxiety Inventory (Spielberger et al., 1970). Scores from other
measures were also included in each symptom
domain if the alternative measure adequately
represented the domain. For example, single flashback or nightmare scores were not included in the
intrusion score as they represent only a portion of
the domain of intrusion symptoms.
The table also shows that scores for observerrated intrusions, avoidance, and total PTSD symptoms were typically obtained from the Clinician
Administered PTSD Scale (Blake et al., 1995) or from
the Structured Interview for PTSD (Davidson et al.,
1989). Observer-rated depression was obtained
from the Hamilton Rating Scale for Depression
(Hamilton, 1960), and observer-rated anxiety was
typically assessed by from the Hamilton Rating
Scale for Anxiety (Hamilton, 1959).
RESULTS
Preliminary Analyses
A series of preliminary analyses were conducted to
determine whether it was necessary to match
# 1998 John Wiley & Sons, Ltd.
133
treatment conditions on important variables
(e.g. treatment duration) before comparing effect
sizes for treatment outcome. In the aggregate,
treatment and control conditions (as listed in
Table 2) did not differ in the mean duration of
their trials, F(13, 47) ˆ 1.86, p 4 0.05 (M ˆ 8.2
weeks, SD ˆ 4.7 weeks). Note, however, that
EMDR trials tended to be shorter than those of
behaviour therapy (M ˆ 3.7 versus 10.1 weeks,
respectively; t(22) ˆ 4.66, p 5 0.001), and consisted
of fewer treatment sessions (M ˆ 4.6 versus 14.8,
respectively; t(22) ˆ 5.51, p 5 0.001). We will return
to consider these differences later in this article.
The effect sizes of studies that included a
control group were compared to effect sizes of
uncontrolled studies. There was no significant
difference for either self-report, F(1, 40) ˆ 2.77,
p 4 0.1, or observer-rated measures of total PTSD
symptoms, F(1, 32) ˆ 0.62, p 4 0.1. Among the
psychological therapies, 75% reported the level of
therapist training. Studies were coded as having
adequate therapist training if they specifically
reported adequate years of therapist experience
(e.g. over 5 years) or formal training with a senior
colleague experienced in the treatment modality. In
the aggregate (i.e. across treatment conditions),
effect sizes did not significantly differ on either
self-report or observer-rated measures for trials that
reported therapist training versus those that did not
report on this variable, ps 4 0.1.
Three types of trauma were classified: combatrelated, rape or assault-related, and a category
reflecting a mix of various trauma or another
trauma. Across the 59 trials that reported trauma
type, 51% involved combat-related trauma only,
19% rape or assault-related trauma only, and 30% a
mix of trauma or other trauma. Within each
treatment condition (for conditions with three or
more trials), mean effect sizes did not significantly
differ across trauma types, ps 4 0.1.
We intended to examine the relationship between
degree of psychiatric comorbidity and effect size.
However, there were insufficient data for such
analysis. Of the 61 trials, only 21% reported a
quantifiable level of comorbid anxiety disorders
apart from PTSD. A total of 33% of trials reported
comorbidity data on mood disorders, and 16% of
trials reported comorbidity data for substance use
disorders. Thus, we did not examine the relationship between effect size and comorbidity.
Follow-up data were not reported for drug
therapies. For the control conditions, follow-up
data were reported for only one supportive therapy
trial. For several psychological therapies, follow-up
Clin. Psychol. Psychother. 5, 126±144 (1998)
134
data were available only on few outcome measures.
Only behaviour therapy and EMDR provided
sufficient follow-up data. Analysis of follow-up
data was therefore restricted to these two treatment
conditions. Across these conditions, the duration
between posttreatment and follow-up did not differ
significantly, F(1, 17) ˆ 1.68, p 4 0.1 (grand M ˆ 15
weeks; behaviour therapy: M ˆ 18 weeks, SD ˆ 12;
EMDR: M ˆ 12 weeks, SD ˆ 6).
Comparison of Treatments
Data Analysis
Rather than conducting multiple comparisons
(e.g. t-tests) between the 14 treatment conditions,
the simplest and most efficient method of evaluating the comparative efficacy of treatments is to
compute confidence intervals. Given the small
number of trials in each condition for each outcome
measure, we chose to use 90th percentile confidence
intervals. Confidence intervals have methodological
advantages over the conventional use of p-values
(see Cohen, 1990). Accordingly, 90% confidence
intervals were calculated for all conditions across all
measures. Conditions with non-overlapping confidence intervals differ significantly at p 5 0.10. Some
treatment conditions consisted of only a single trial,
in which case it is not possible to compute
confidence intervals. However, the effect sizes of
these trials could be compared to the confidence
intervals of other trials, to determine whether the
obtained effect size fell within the confidence
interval. If the effect size falls within the interval,
then the two conditions do not significantly differ.
For each dependent measure, comparisons were
as follows: (1) general comparisons between the
overall drug therapy, psychological therapy, and
control groups; (2) comparison of treatments to
controls and to one another (e.g. SSRIs versus
controls; SSRIs versus TCAs); (3) comparisons
across treatment types (e.g. SSRIs versus EMDR);
(4) where relevant, comparisons between control
conditions (e.g. WLCs and pill placebo versus
supportive psychotherapy). Following the recommendations of Hunter and Schmidt (1990), Wolf
(1986), and others, we made these comparisons by
computing weighted means. That is, means were
computed by weighting the effect size of each trial
by the number of participants completing that trial.
This procedure gives greater weighting to the effect
sizes of larger trials, which are likely to be more
reliable estimates of treatment efficacy than the
effect sizes of small trials (Hunter and Schmidt,
# 1998 John Wiley & Sons, Ltd.
M. L. Van Etten and S. Taylor
1990). Confidence intervals were computed around
weighted means.
Attrition
Table 2 shows the proportions of dropouts for
each treatment condition. With regard to the three
classes of treatment conditions (i.e. drugs treatments, psychological therapies, and control conditions), the table shows that attrition was
significantly greater (i.e. confidence intervals did
not overlap) for drug therapies (M ˆ 31.9%) compared to psychological therapies (14.0%) and controls (16.6%). Within the drug therapy conditions,
the dropout rates tended not to differ from one
another. No differences in dropout rates were
observed within the psychological therapy conditions or the control conditions, or between the
psychological therapies and controls.
Effect Sizes at Posttreatment: PTSD Symptoms
Intrusions (Self-report)
Table 2 shows that for the three classes of
treatment conditions (drugs treatments, psychological therapies, and control conditions), drug
therapies and psychological therapies were generally equally effective (i.e. their confidence intervals
overlapped), and only psychological therapies were
significantly superior to controls. Among the drug
therapies, all treatments were superior to WLCs.
Both SSRIs and carbamazepine were superior to
supportive therapy controls. Only the effect size for
the single carbamazepine was significantly greater
than that of pill placebo. SSRIs and carbamazepine
had comparable effect sizes, and both were superior
to all other drug therapies.
All psychological therapies were significantly
superior to WLCs, and none were significantly
better than pill placebo. Behaviour therapy was as
effective as other psychological therapies, and
EMDR was superior to relaxation and dynamic
therapy. Across drug therapies and psychological
therapies, SSRIs were superior to all psychological therapies except EMDR and behaviour
therapy. Carbamazepine was similarly superior
to all psychological therapies except behaviour
therapy.
Intrusions (Observer-rated)
Only some drug therapies (one TCA trial, SSRIs,
and one BDZ trial), two of the psychological
therapies (behaviour therapy and EMDR), and
two control conditions (WLCs and one supportive
psychotherapy) reported results for observer-rated
Clin. Psychol. Psychother. 5, 126±144 (1998)
Condition
TCA
Carbmz
MAOI
SSRI
BDZ
No.
trials
% dropout
M
90%CI
Intrusions
Self-report
Observer-rated
M
90%CI
M
90%CI
Avoidance
Self-report
Observer-rated
M
90%CI
M
90%CI
Total severity of PTSD symptoms
Self-report
Observer-rated
M
90%CI
M
90%CI
26.4
9.0
36.4
36.0
38.0
14.4±38.4
Ð
24.7±48.2
6.5±65.5
Ð
0.64
1.53
0.64
1.71
0.51
0.30±0.98
Ð
0.27±1.01
1.08±2.34
Ð
0.46
Ð
Ð
1.28
0.66
Ð
Ð
Ð
0.90±1.66
Ð
0.35
0.52
0.40
0.92
0.16
0.22±0.48
Ð
ÿ0.21±1.01
0.73±1.11
Ð
0.55
Ð
Ð
1.37
0.32
Ð
Ð
Ð
1.05±1.69
Ð
0.54
0.93
0.61
1.38
0.49
0.34±0.74
Ð
0.38±0.84
1.02±1.74
Ð
0.86
1.45
0.92
1.43
0.54
0.75±0.97
Ð
0.73±1.11
1.19±1.67
Ð
Drug Txs
(overall)
19
31.9
25.4±38.4
0.86
0.63±1.09
1.01
0.71±1.31
0.45
0.31±0.59
1.00
0.64±1.36
0.69
0.55±0.83
1.05
0.91±1.19
Behav Tx
EMDR
Relaxat'n
Hypnosis
Dynamic
13
11
1
1
1
15.1
14.4
8.0
11.0
11.0
9.8±20.4
7.8±21.0
Ð
Ð
Ð
1.12
1.12
0.54
1.06
0.70
0.49±1.75
0.72±1.52
Ð
Ð
Ð
1.76
1.39
Ð
Ð
Ð
ÿ0.05±3.57
0.99±1.79
Ð
Ð
Ð
1.12
1.27
0.46
0.80
0.64
0.61±1.63
0.74±1.80
Ð
Ð
Ð
1.45
2.01
Ð
Ð
Ð
ÿ0.10±3.00
1.25±2.77
Ð
Ð
Ð
1.27
1.24
0.45
0.94
0.90
0.80±1.74
0.99±1.49
Ð
Ð
Ð
1.89
0.69
Ð
Ð
Ð
1.66±2.12
ÿ0.06±1.44
Ð
Ð
Ð
Psych Tx
(overall)
27
14.0
10.8±17.2
1.02
0.80±1.24
1.57
1.12±2.02
1.03
0.77±1.29
1.74
1.23±2.25
1.17
0.99±1.35
1.51
1.17±1.85
Pill Plac
WLC
Sup Psych
No Sacc
4
5
5
1
23.0
6.2
20.5
Ð
6.6±39.4
0.2±12.2
8.5±32.5
Ð
0.48
0.32
0.95
Ð
ÿ0.17±1.13
0.28±0.36
Ð
Ð
Ð
0.74
0.53
Ð
Ð
0.72±0.76
Ð
Ð
0.07
0.21
0.77
Ð
0.05±0.09
0.14±0.28
Ð
Ð
Ð
0.22
0.09
Ð
Ð
ÿ0.65±1.09
Ð
Ð
0.51
0.44
0.34
0.22
0.29±0.73
0.28±0.60
0.01±0.67
Ð
0.77
0.75
0.92
Ð
0.63±0.91
0.67±0.83
Ð
Ð
Controls
(overall)
15
16.6
10.5±22.7
0.49
0.29±0.69
0.66
0.54±0.78
0.23
0.06±0.46
0.17
ÿ0.18±0.52
0.43
0.33±0.53
0.77
0.71±0.83
p
Effect size ˆ (Mpre 7 Mpost)/SDpooled , where SDpooled ˆ ‰…SD2pre ‡ SD2post †=2Š. All means are weighted by sample size. See text for details. 90%CI ˆ 90th percentile confidence interval
arounded weighted mean. Note that `Ð' refers to data missing or not reported. For the 90%CIs `Ð' appears when there was only one effect size. Within each row, total number of
trials may differ across outcome domains (intrusions, avoidance, and global severity) because some trials did not assess all domains.
BDZ, benzodiazepines; Behav Tx, behaviour therapy; Carbmz, carbamazepine; Dynamic, psychodynamic psychotherapy; EMDR, eye movement desensitization and reprocessing;
MAOI, monoamine oxidase inhibitors; No Sacc, no saccade control (control for EMDR); Pill Plac, pill placebo; Relaxat'n, relaxation training; SSRI, selective serotonin reuptake
inhibitors; Sup Psych, supportive psychotherapy; TCA, tricyclic antidepressants; WLC, waiting-list control.
135
Clin. Psychol. Psychother. 5, 126±144 (1998)
6
1
7
4
1
Meta-Analysis of PTSD Treatments
# 1998 John Wiley & Sons, Ltd.
Table 2. Dropout proportions and pre±post effect sizes for measures of PTSD symptoms
136
measures. Among these trials, the drug therapies
and the psychological therapies were generally
equally effective, although there was a trend for
psychological therapies to have larger effect sizes
(Table 2). Drug and psychological therapies were
more effective than controls. Within the drug
therapies, only SSRIs were more effective than all
controls, and SSRIs were also more effective than all
other drug therapies. Within the psychological
therapies, EMDR and behaviour therapy demonstrated comparable effect sizes, but only EMDR was
significantly more effective than all controls. EMDR
was more effective than the most effective drug
therapy, the SSRIs.
Avoidance (Self-report)
Table 2 shows that psychological therapies were
more effective than both drug therapies and control
conditions. All drug therapies were more effective
than pill placebo and WLCs, and none were more
effective than the single supportive therapy control.
SSRIs were significantly more effective than all
other drug therapies. All psychological therapies
were more effective than pill placebo and WLCs,
but again, not more effective than the single
supportive psychotherapy trial. Among the psychological therapies, behaviour therapy and EMDR
were equally effective, but only EMDR was
superior to relaxation and dynamic therapy.
SSRIs, EMDR, behaviour therapy and hypnotherapy did not differ in effect sizes.
Avoidance (Observer-rated)
Only some of the drug therapies (one TCA trial,
SSRIs, and one BDZ trial), two of the psychological
therapies (behaviour therapy and EMDR), and two
control conditions (WLCs and one supportive
psychotherapy) reported on avoidance. Drug therapies and psychological therapies were more effective than controls, and were equally effective to one
another (Table 2). However, there was a trend for
psychological therapies to have a larger effect size.
Among the drug therapies, all were significantly
more effective than supportive psychotherapy, and
there was a trend for SSRIs to also be more effective
than WLCs. Among the psychological therapies,
EMDR and behaviour therapy were equally effective, but only EMDR was more effective than the
control conditions. SSRIs, behaviour therapy, and
EMDR were all equally effective.
Total PTSD Symptoms (Self-report)
As seen in Table 2, psychological therapies were
more effective than drug therapies, and both were
# 1998 John Wiley & Sons, Ltd.
M. L. Van Etten and S. Taylor
more effective than controls. Among the drug
therapies, only the SSRIs and carbamazepine were
more effective than all control conditions. The SSRIs
had a significantly greater mean effect size than all
other drug therapies. Among the psychological
therapies, all but relaxation were significantly
more effective than all control groups. EMDR and
behaviour therapy were equally effective, but only
EMDR was significantly superior to all other
psychological therapies. Behaviour therapy and
EMDR were equally effective as the SSRIs.
Total PTSD Symptoms (Observer-rated)
As seen in Table 2, outcome data for total PTSD
symptoms were reported by all drug therapies
and all controls except the non-saccade control.
However, these data were reported by only two
psychological therapies (behaviour therapy and
EMDR). Psychological therapies and drug therapies
were equally effective, although there was a trend
for psychological therapies to have larger effect
sizes. Drug and psychological therapies were
more effective than controls. Among the drug
therapies, only the SSRIs and carbamazepine were
more effective than all controls, and both were also
more effective than all other drug therapies. Among
the psychological therapies, only behaviour therapy
was more effective than all controls, and behaviour
therapy was also more effective than EMDR. The
best psychological therapy, behaviour therapy, was
significantly more effective than carbamazepine,
and there was a trend for behaviour therapy to be
more effective than SSRIs as well.
Effect Sizes at Posttreatment:
Anxiety and Depression
Anxiety (Self-report)
Psychological therapies were more effective than
drug therapies, and both were more effective than
controls (Table 3). All psychological therapies were
more effective than all controls. All drug therapies
except TCAs were superior to all controls. The SSRI
trial tended to be superior to all other drug
therapies. Within the psychological therapies, the
largest effect sizes were observed in the EMDR and
behaviour therapy conditions. Of these conditions,
only behaviour therapy was more effective than
relaxation therapy. Across drug therapies and
psychotherapies, SSRIs and behaviour therapy
were equal in efficacy. The SSRI trial was comparable to behaviour therapy, but more effective than
EMDR and other psychological therapies. EMDR
Clin. Psychol. Psychother. 5, 126±144 (1998)
Meta-Analysis of PTSD Treatments
137
Table 3. Pre±post effect sizes for measures of anxiety and depression
Condition
Anxiety
Self-report
M
90%CI
Observer-rated
M
90%CI
Depression
Self-report
Observer-rated
M
90%CI
M
90%CI
TCA
Carbmz
MAOI
SSRI
BDZ
Drug Txs
(overall)
0.44
0.47
0.65
1.24
Ð
0.61
ÿ0.08±0.96
Ð
Ð
Ð
Ð
0.39±0.83
0.54
1.73
0.92
1.20
0.72
0.64
0.13±0.95
Ð
0.44±1.40
Ð
Ð
0.61±1.09
0.44
0.48
0.98
1.41
Ð
0.65
0.09±0.79
Ð
Ð
Ð
Ð
0.39±0.91
0.85
1.25
0.43
1.38
0.11
0.72
0.53±1.17
Ð
0.28±0.58
Ð
Ð
0.55±0.89
Behav Tx
EMDR
Relaxat'n
Hypnosis
Dynamic
Psych Tx
(overall)
1.12
0.95
0.83
0.95
1.07
1.04
0.84±1.40
0.69±1.21
Ð
Ð
Ð
0.89±1.19
1.47
Ð
Ð
Ð
Ð
1.47
Ð
Ð
Ð
Ð
Ð
Ð
0.97
1.05
0.67
Ð
Ð
1.00
0.80±1.14
0.81±1.29
Ð
Ð
Ð
0.87±1.13
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Pill Plac
WLC
Sup Psych
No Sacc
Controls
(overall)
0.03
0.25
0.25
0.06
0.17
Ð
0.14±0.36
0.04±0.46
Ð
0.06±0.28
0.38
Ð
Ð
Ð
0.38
Ð
Ð
Ð
Ð
Ð
0.24
0.25
0.22
0.14
0.23
Ð
0.12±0.42
0.07±0.37
Ð
0.16±0.30
0.36
Ð
Ð
Ð
0.36
0.19±0.53
Ð
Ð
Ð
0.19±0.53
See footnote to Table 2 for a definition of statistics and acronyms.
yielded a larger effect size than the non-saccade
control.
Anxiety (Observer-rated)
As seen in Table 3, among the psychological
therapies, only one trial of behaviour therapy
reported on this measure, and only one pill placebo
trial was available among controls. The behaviour
therapy trial demonstrated a significantly greater
effectiveness than the drug therapies, and both were
more effective than pill placebo. Among the drug
therapies, the carbamazepine trial and SSRI trial
were more effective than all other drug therapies.
Depression (Self-report)
Table 3 shows that the drug therapies and
psychological therapies were generally comparable
in efficacy, although there was a trend for psychological therapies to have larger effect sizes. Both
were more effective than the control conditions. All
drug therapies except TCAs were superior to all
controls. The one SSRI trial tended to be more
effective than all other drug therapies. All psychological therapies were more effective than all control
conditions. Behaviour therapy and EMDR were
equally effective, and both were more effective than
# 1998 John Wiley & Sons, Ltd.
the relaxation condition. The single SSRI trial was
significantly more effective than the best psychological therapies, EMDR and behaviour therapy.
Depression (Observer-rated)
No psychological therapies reported observerrated depression, and only two pill placebo trials
reported outcomes among the control conditions.
All drug therapies except the BDZ trial were more
effective than pill placebo. Among the drug
therapies, the SSRIs were more effective than
carbamazepine, and both SSRIs and carbamazepine
had larger effect sizes than MAOIs and TCAs.
Posttreatment: Summary of Results
Psychological therapies tended to be more effective
than drug therapies, and both tended to be more
effective than controls. Of the drug therapies,
the SSRIs and carbamazepine were the most
effective, although the carbamazepine condition
was based on only a single trial. SSRIs tended to
be more effective in treating intrusions than
avoidance symptoms according to self-report but
not observer-rated measures. Other drug therapies
generally demonstrated small effect sizes in relation
Clin. Psychol. Psychother. 5, 126±144 (1998)
138
M. L. Van Etten and S. Taylor
to control conditions, and affected only a few of the
symptom domains. Of the psychological therapies,
behaviour therapy and EMDR were the most
effective, with the two being generally equally
efficacious, although behaviour therapy was significantly more effective than all treatments, including
EMDR, SSRIs, and carbamazepine, on observerrated total PTSD symptoms. No differences in
comparative treatment efficacy were discernible
between behaviour therapy and EMDR across the
specific symptom domains of PTSD.
The less effective psychological therapies demonstrated moderate yet consistent effect sizes across
symptom domains in relation to control conditions,
with relaxation being the least effective, followed by
dynamic therapy and lastly, hypnotherapy. However, these conditions were all based upon single
trials, and therefore should be interpreted with
caution. Supportive psychotherapies tended to
demonstrate efficacy comparable to the less effective psychological therapies and drug therapies.
The best psychological therapies, behaviour
therapy and EMDR, tended to be as effective as the
best drug therapies, SSRIs and carbamazepine,
across PTSD outcomes domains. However, as noted
above, behaviour therapy was more effective than all
treatments on observer-rated total PTSD symptoms.
SSRIs may have been more effective than behaviour
therapy and EMDR in treating depression, but this
was based upon a single SSRI trial, and results were
only available for self-reported depression.
Effect Sizes at Follow-up
No drug therapy, pill placebo, or wait-list control
results were available for follow-up, and only single
trials of other psychological therapies and the
supportive psychotherapy trial provided follow-up
data, and only across some symptom measures.
Therefore the only conditions which provided
adequate trials for calculating follow-up effect
sizes across most symptom domains were behaviour therapy and EMDR. Table 4 shows the followup effect sizes for PTSD symptoms, and Table 5
shows the results for measures of anxiety and
depression. As noted in the section titled `Preliminary analyses', the mean duration between posttreatment and follow-up did not differ significantly
across conditions (mean follow-up duration ˆ 15
weeks).
Across all self-report and observer-rated measures
of PTSD symptoms, depression and anxiety, both
behaviour therapy and EMDR demonstrated a
maintenance of treatment effects at follow-up,
with no differences between the two conditions at
follow-up on any measures. Differences in effect size
from posttreatment to follow-up were nonsignificant
for all measures across both conditions, except that
EMDR demonstrated a significant increase in effect
size for observer-rated total PTSD symptoms at
follow-up, making it equal to behaviour therapy,
whereas EMDR was less effective than behaviour
therapy for this measure at posttreatment.
Table 4. Effect sizes at follow-up (i.e. symptom reductions from pretreatment to 15 week follow-up) for PTSD
symptoms
Condition
No.
trials
Behav Tx
EMDR
5
6
Intrusions
Self-report
Observer-rated
M
90%CI
M
90%CI
1.56
1.75
0.81±2.29
1.46±2.04
1.47
2.07
0.60±2.34
1.77±2.37
Avoidance
Self-report
Observer-rated
M
90%CI
M
90%CI
1.44
1.89
0.47±2.41
1.08±2.70
1.32
2.34
0.71±1.93
1.76±2.92
Total severity of PTSD symptoms
Self-report
Observer-rated
M
90%CI
M
90%CI
1.63
1.33
1.10±2.16
0.89±1.77
1.93
2.27
1.67±2.19
1.78±2.76
See footnote to Table 2 for a definition of statistics and acronyms.
Table 5. Effect sizes at follow-up for measures of anxiety and depression
Condition
No. trials
Behav Tx
EMDR
9
5
Anxiety
Self-report
M
90%CI
0.99
0.90
Depression
Self-report
M
90%CI
0.66±1.32
0.64±1.16
0.93
0.91
0.76±1.10
0.46±1.36
See footnote to Table 2 for a definition of statistics and acronyms.
# 1998 John Wiley & Sons, Ltd.
Clin. Psychol. Psychother. 5, 126±144 (1998)
Meta-Analysis of PTSD Treatments
Fail-Safe N
Recall that fail-safe N is defined as the number of
unpublished null trials (i.e. those obtaining zero
effect sizes) required to reduce an obtained mean
effect size to a trivial magnitude (0.050). An average
of self-reported and observer-rated total PTSD
symptom effect sizes were used for dk (mean
obtained effect size) in calculating fail-safe N
calculations. For the posttreatment data, the necessary number of unpublished null trials in each active
treatment condition were as follows: TCA (41),
carbamazapine (23), MAOI (66), SSRI (95), BDZ
(10), behaviour therapy (220), EMDR (139), relaxation therapy (8), hypnotherapy (18), and dynamic
therapy (17). These results suggest that for most
conditions, there would need to be a large number
of unpublished null trials to reduce obtained mean
effect sizes to trivial levels. It seems unlikely,
especially for SSRIs, EMDR, and behaviour therapy,
that there would be so many unpublished trials
finding zero effects for these treatments, and so we
conclude that these findings were unlikely to have
been biased by the `file-drawer' problem. However,
significantly fewer unpublished null trials were
indicated to suggest bias in results for the carbamazepine, BDZ, relaxation, hypnotherapy, and
dynamic therapy conditions. Therefore these conditions should be viewed cautiously as they may be
subject to the file-drawer bias.
DISCUSSION
The purpose of the present study was to empirically
evaluate the relative efficacy of various treatments
for PTSD. We found that behaviour therapy and
EMDR were the most effective psychological
therapies for PTSD. Effect sizes for these therapies
were large relative to control conditions, indicating
strong treatment effects, and dropout rates were
low, indicating good treatment acceptance. The
SSRIs and carbamazepine were the most effective
drug therapies, with similarly large effect sizes, but
dropout was fairly high (Table 2). Even SSRIs,
characteristically more acceptable to patients due to
a more tolerable side-effect profile, were tolerated
poorly by the PTSD patients. An average of 36% of
patients treated with SSRIs discontinued treatment prematurely. It may be that PTSD patients,
characteristically hyperaroused, may be particularly
sensitive to side-effects occasionally associated with
SSRIs (e.g. agitation), and may discontinue secondary to these effects.
# 1998 John Wiley & Sons, Ltd.
139
Although dropout rates were high for drug
therapies, the SSRIs demonstrated the greatest
treatment efficacy of all drug therapies (other than
carbamazepine, for which there was only one trial),
with large effect sizes apparent across all symptom
domains. Effects were perhaps stronger for intrusive symptoms and depressive symptoms than for
avoidance symptoms, but all effect sizes were large
relative to control conditions. However, posttreatment assessments for drug therapies occurred prior
to medication discontinuation, so it is not clear
whether treatment effects maintain when medications are withdrawn.
These results suggest that although SSRIs may
not be the treatment of choice for PTSD, given the
higher dropout rate, they may still be an acceptable
and useful treatment if the patient can be retained
in treatment, and may be particularly useful for
patients whose intrusive and depressive symptoms predominate their experience. Unfortunately,
follow-up data were unavailable for these trials, so
it is unclear as to whether effects were maintained
over time, or even immediately after medication
discontinuation. Further research is warranted to
investigate the maintenance of SSRI effects in PTSD
patients. Carbamazepine also appeared to be an
effective treatment for PTSD, but results were based
upon a single trial and should therefore be
interpreted with caution.
Only one BDZ trial was available for this metaanalysis. This study used alprazolam at a moderately high dose (1.5 to 6 mg/day) and generally
failed to demonstrate treatment efficacy (Braun et al.,
1990). Effect sizes were small if not trivial relative to
control conditions. Although not included in our
meta-analysis due to failure to meet our 100% PTSD
diagnosis rule, another study also supports the
inefficacy of BDZs for PTSD. Gelpin et al. (1996)
reported that clonazepam and alprazolam failed to
benefit acute trauma survivors' PTSD and anxiety
symptoms over controls. In another study not
included in the meta-analysis due to lack of a
standard posttreatment assessment time across
patients, alprazolam did not alter auditory startle
response in PTSD patients (Bloch et al., 1996). This is
enlightening given the high frequency of BDZ
prescriptions by general practitioners for anxiety
disorders. If medications are to be prescribed for
PTSD treatment, it appears as though SSRIs are a
more effective treatment than BDZs. The more
narrow side-effect profile, non-addictiveness, and
greater overdose threshold of SSRIs over BDZs also
support this point (Canadian Pharmaceutical
Association, 1996).
Clin. Psychol. Psychother. 5, 126±144 (1998)
140
As noted, behaviour therapy and EMDR were the
most effective psychological therapies, and both
were as effective as SSRIs. Effect sizes were large
across all PTSD symptom domains for these
treatments in relation to controls, and treatments
were generally statistically comparable in efficacy,
with some minor exceptions. For example, behaviour therapy was more effective than EMDR and
SSRIs on observer-rated total PTSD symptoms at
posttreatment. However, by follow-up, the differences between behaviour therapy and EMDR were
nonsignificant (SSRIs did not report follow-up
data). Because attrition from posttreatment to
follow-up was comparable across EMDR and
behaviour therapy conditions, the `catching up' of
EMDR to behaviour therapy cannot be attributed to
differential attrition of poor responders. No
differences in treatment efficacy between behaviour
therapy and EMDR were noted across the specific
PTSD symptom domains. Both treatments maintained effects at follow-up, and were equally
effective on all measures at follow-up.
The efficacy of EMDR in PTSD treatment is a
controversial finding in light of recent discussions in
the literature critically questioning the validity of
EMDR as a treatment for anxiety disorders (Herbert
and Mueser, 1992; Lohr et al., 1995; Lilienfeld, 1996).
The results of the present study suggest that EMDR
is effective for PTSD, and that it is more efficient
than other treatments. Despite its apparent efficacy,
what works in EMDR and the mechanism for how
it works remains unclear. That is, we know little
about the active ingredients in EMDR and the
mechanisms by which these ingredients result in
decreased PTSD symptoms. EMDR theory suggests
that eye movements or other oscillatory movements during trauma imagining somehow result in
brain alterations which then allow more appropriate
processing of the trauma, thereby reducing PTSD
symptoms. This still does not inform us of what
changes occur in what part of the brain, how
oscillatory movements are involved in those
changes, how that leads to `reprocessing' of the
trauma, and how such reprocessing results in
decreased PTSD symptoms. Some might argue
that EMDR works through exposure and desensitization, similar to behaviour therapy. However, this
is unlikely to be the case given that EMDR provides
significantly less trauma exposure than behaviour
therapy and is demonstrating comparable effects,
which suggests that another treatment component
specific to EMDR is active. Expectancy may play a
role, but this has not been thoroughly examined,
and is inconsistent with our finding that the effect
# 1998 John Wiley & Sons, Ltd.
M. L. Van Etten and S. Taylor
sizes of EMDR tended to be larger than those of
control conditions, such as pill placebo and supportive psychotherapy.
Regarding the question of active ingredients in
EMDR, even the utility of the actual eye-movements
in EMDR is unclear, since other stimuli are used
clinically. Only one non-saccade trial was available
for comparison with the EMDR trials in our metaanalysis (Devilly and Spence, 1996), therefore
permitting little comment on this issue. When all
eye-movement conditions in the meta-analysis were
compared to this one fixed-eye condition, EMDR
was more effective than the fixed-eye control across
measures. However, when the fixed-eye control was
compared to the EMDR condition within the same
study (i.e. Devilly and Spence, 1996), the fixed-eye
condition was comparable to EMDR across
measures. However, that study may have been
compromised by insufficient treatment duration
because only two EMDR sessions for multiply
traumatized sessions were used. Thus, there may
not have been sufficient time to treat each traumatic
memory.
Pitman and colleagues (1996b) also included a
fixed-eye control (that included hand-tapping) in
their EMDR study. This was not included in the
meta-analysis because it was tested within a
crossover design for which we already included
the EMDR treatment phase. Pitman and colleagues
report that the fixed-eye condition performed as
well if not better than the EMDR trial. Another
study not included in the meta-analysis (due to the
fact that not all patients had full PTSD) reported
that a fixed-eye control demonstrated large treatment effect sizes comparable to that of the EMDR
condition, although the eye movement condition
was more efficient (Renfrey and Spates, 1994).
This study also demonstrated that a condition
using light bars versus a therapist's finger to
stimulate eye-movements was also comparable to
the other conditions, suggesting that alternate
means of eye-movement production may be comparable to the procedures originally described for
EMDR. Clarification of the mechanisms by which
symptoms change and the active ingredients in
EMDR is now critical given its apparent efficacy.
Without such clarification, the acceptability of
EMDR within the professional community is likely
to remain controversial.
Recent efforts in treatment outcome research have
been made to examine both separate and combined
treatments for various disorders. In a recent metaanalysis of treatments for panic disorder, Gould
et al. (1995) reported that the combination of
Clin. Psychol. Psychother. 5, 126±144 (1998)
Meta-Analysis of PTSD Treatments
benzodiazepine treatment with behaviour therapy
is less effective than behaviour therapy alone.
Future research is needed to determine whether a
similar finding holds for PTSD. That is, whether
combined benzodiazepine therapy and psychological therapy is less effective than the latter
alone. Another direction for future research
suggested by the results of this meta-analysis is
the study of the effectiveness of combined psychological therapies (e.g. combined behaviour therapy
and EMDR) versus individual psychological therapies. Finally, we focused our analysis on studies of
fully diagnosed chronic PTSD in adults. Future
studies examining the comparability of treatment
efficacy in subclinical PTSD and full PTSD in
children would also enhance our understanding of
the treatment of psychological trauma.
A limitation of this paper is the relatively low
number of trials available per treatment condition,
and our corresponding use of 90% rather than 95%
confidence intervals. Yet despite the small number
of trials, the number of trials was sufficient to detect
differences in effect sizes across various treatment
conditions. Even so, more research trials on PTSD
treatment outcome would likely increase the power
of the meta-analysis and would also permit more
fine-grained analyses. Future research should
address this issue. Another limitation that was
true for our study and is also common to many
meta-analyses is that we were only able to base our
results on completer-analyses. It remains to be seen
whether our conclusions hold up for intent-to-treat
analyses. Although effect sizes should be smaller
for intent-to-treat analyses, we expect that most
of these findings would maintain. One exception
is that the effect sizes for drug treatments
may diminish more than those for psychological
therapies, because the former had more dropouts,
and dropouts may have poorer treatment response.
To conclude, this meta-analysis provides new
information about the relative efficacy of treatments
for PTSD. Our results support the use of behaviour
therapy, EMDR, and SSRIs. It remains to be seen
whether the efficacy of treatment can be improved
by using these interventions in combination.
ACKNOWLEDGEMENTS
This research was supported in part by a grant from
the Medical Research Council of Canada and a
training grant from the National Institute on Drug
Abuse. Thanks to Drs Baker, Blanchard, Burstein,
Carlson, Devilly, Foa, Frueh, Lazrove, Lipper, Lohr,
# 1998 John Wiley & Sons, Ltd.
141
Marcus, Pitman, Rothbaum, Scheck, Shapiro and
Wilson for providing unpublished data.
REFERENCES
Abramowitz, J. S. (1997a). Effectiveness of psychological and pharmacological treatments for obsessivecompulsive disorder: A quantitative review. Journal of
Consulting and Clinical Psychology, 65, 44±52.
Abramowitz, J. S. (1997b). Variants of exposure and
response prevention in the treatment of obsessivecompulsive disorder: A meta-analysis. Behaviour
Therapy, 27, 583±600.
American Psychiatric Association. (1994). Diagnostic
and Statistical Manual of Mental Disorders, 4th edn.
Washington, DC: Author.
Baker, D. G., Diamond, B. I., Gillette, G., Hammer, M.,
Katzelnick, D., Keller, T., Mellman, T. A., Pontius, E.,
Rosenthal, M., Tucker, P., van der Kolk, B. A. and
Katz, R. (1995). A double-blind, randomized, placebocontrolled, multi-center study of brofaromine in the
treatment of post-traumatic stress disorder. Psychopharmacology, 122, 386±389.
Beck, A. T., Rush, A. J., Shaw, B. and Emery, G. (1979).
Cognitive Therapy of Depression. New York: Guilford.
Blake, D. D., Weathers, F. W., Nagy, L. M., Kaloupek,
D. G., Gusman, F. D., Charney, D. S. and Keane, T.
(1995). The development of a clinician-administered
PTSD Scale. Journal of Traumatic Stress, 8, 75±90.
Bloch, M., Peri, T., Bonne, O. and Shalev, A. Y. (1996).
Effect of alprazolam on auditory startle in PTSD and
panic disorder. Unpublished manuscript, Department of
Psychiatry, Hadassah University Hospital, Jerusalem,
Israel.
Braun, P., Greenberg, D., Dasberg, H. and Lerer, B. (1990).
Core symptoms of post-traumatic stress disorder
unimproved by alprazolam treatment. Journal of Clinical
Psychiatry, 51, 236±238.
Breslau, N., Davis, G. C., Andreski, P. and Peterson, E.
(1991). Traumatic events and post-traumatic stress
disorder in an urban population of young adults.
Archives of General Psychiatry, 48, 216±222.
Brom, D., Kleber, R. J. and Defares, P. B. (1989).
Brief Psychotherapy for post-traumatic stress disorders.
Journal of Consulting and Clinical Psychology, 57, 607±612.
Burstein, A. (1984). Treatment of post-traumatic stress
disorder with imipramine. Psychosomatics, 25, 681±687.
Canadian Pharmaceutical Association (1996). Compendium
of Pharmaceuticals and Specialties, 31st edn. Ottawa,
Ontario: Author.
Carlson, J. G., Chemtob, C. M., Rusnak, K., Hedlund, N. L.
and Muraoka, M. Y. (in press). Eye-movement desensitization and reprocessing (EMDR) treatment for combat-related post-traumatic stress disorder. Journal of
Traumatic Stress.
Chemtob, C., Roitblat, H. L., Hamada, R. S., Carlson, J. G.
and Twentyman, C. T. (1988). A cognitive action theory
of post-traumatic stress disorder. Journal of Anxiety
Disorders, 2, 253±275.
Christensen, H., Hadzi-Pavlovic, D., Andrews, G. and
Mattick, R. (1987). Behaviour therapy and tricyclic
Clin. Psychol. Psychother. 5, 126±144 (1998)
142
medication in the treatment of obsessive-compulsive
disorder: A quantitative review. Journal of Consulting
and Clinical Psychology, 55, 701±711.
Coccaro, E. F. and Siever, L. J. (1995). The neuropsychopharmacology of personality disorders. In: F. E. Bloom
and D. J. Kupfer (Eds), Psychopharmacology: The Fourth
Generation of Progress. New York: Raven, pp. 1567±1579.
Cohen, J. (1988). Statistical Power Analysis for the Behavioural Sciences, 2nd edn. Hillsdale, NJ: Erlbaum.
Cohen, J. (1990). Things I have learned (so far). American
Psychologist, 45, 1304±1312.
Cooper, N. A. and Clum, G. A. (1989). Imaginal flooding
as a supplementary treatment for PTSD in combat
veterans: A controlled study. Behaviour Therapy, 20,
381±391.
Davidson, J. (1987). A pilot study of phenelzine in the
treatment of post-traumatic stress disorder. British
Journal of Psychiatry, 150, 252±255.
Davidson, J., Foa, E. B., Blank, A. S., Brett, E. A.,
Fairbank, J., Green, B. L., Herman, J. L., Keane, T. M.,
Kilpatrick, D. L., March, J. S., McNally, R. J., Pitman,
R. K., Resnick, H. S. and Rothbaum, B. O. (1996). Posttraumatic stress disorder. In: T. A. Widiger, A. J. Frances,
H. A. Pincus, R. Ross, M. B. First and W. W. Davis
(Eds), DSM-IV Sourcebook, Vol. 2. Washington, DC:
American Psychiatric Association, pp. 577±605.
Davidson, J. R. T., Hughes, D., Blazer, D. G. and George,
L. G. (1991). Post-traumatic stress disorder in the
community: An epidemiological study. Psychological
Medicine, 21, 713±721.
Davidson, J., Kudler, H., Smith, R., Mahorney, S. L.,
Lipper, S., Hammett, E., Saunders, W. B. and Cavenar,
J. O. (1990). Treatment of post-traumatic stress disorder
with amitriptyline and placebo. Archives of General
Psychiatry, 47, 259±266.
Davidson, J., Smith, R. and Kudler, H. (1989). Validity
and reliability of the DSM-III criteria for post-traumatic
stress disorder: Experience with a structured interview.
Journal of Nervous and Mental Disease, 177, 336±341.
Devilly, G. J. and Spence, S. (1996). The clinical
and statistical efficacy of eye movement desensitization
and reprocessing: Treating PTSD within a veteran
population. Unpublished manuscript, Department of
Psychology, University of Queensland, Australia.
Foa, E. B. (1994, July). Psychopathology and treatment of
PTSD in rape victims. Paper presented at the 102nd
Annual Convention of the American Psychological
Society, Washington, DC.
Foa, E. B. and Kozak, M. J. (1986). Emotional processing
of fear: Exposure to corrective information. Psychological
Bulletin, 99, 20±35.
Foa, E. B., Freund, B. F., Hembree, E., Dancu, C. V.,
Franklin, M. E., Perry, K. J., Riggs, D. S. and Molnar,
C. (1996). Efficacy of short term behavioural treatments
of PTSD in sexual and nonsexual assault victims.
Unpublished manuscript, Medical College of Pennsylvania at Eastern Pennsylvania Psychiatric Institute,
Philadelphia, PA.
Foa, E. B., Rothbaum, B. O., Riggs, D. S. and
Murdock, T. B. (1991). Treatment of post-traumatic
stress disorder in rape victims: A comparison between
cognitive-behavioural procedures and counseling.
# 1998 John Wiley & Sons, Ltd.
M. L. Van Etten and S. Taylor
Journal of Consulting and Clinical Psychology, 59,
715±723.
Foa, E. B., Steketee, G. and Rothbaum, B. O. (1989).
Behavioural/cognitive conceptualizations of posttraumatic stress disorder. Behaviour Therapy, 20,
155±176.
Foa, E. B., Zinbarg, R. and Rothbaum, B. O. (1992).
Uncontrollability and unpredictability in post-traumatic
stress disorder: An animal model. Psychological Bulletin,
112, 218±238.
Forbes, D., Creamer, M. and Rycroft, P. (1994). Eye
movement desensitization and reprocessing in posttraumatic stress disorder: A pilot study using assessment measures. Journal of Behaviour Therapy and
Experimental Psychiatry, 25, 113±120.
Friedman, J. J. (1991). Biological approaches to the
diagnosis and treatment of post-traumatic stress disorder. Journal of Traumatic Stress, 4, 67±91.
Frueh, B. C., Turner, S. M., Beidel, D. C., Mirabella, R. F. and
Jones, W. J. (1996). Trauma management therapy: A
preliminary evaluation of a multicomponent behavioral
treatment for chronic combat-related PTSD. Behaviour
Research and Therapy, 34, 533±543.
Gelpin, E., Bonne, O., Peri, T., Brandes, D. and Shalev, A. Y.
(1996). Treatment of recent trauma survivors with
benzodiazepines: A prospective study. Unpublished
manuscript, Department of Psychiatry, Hadassah University Hospital, Jerusalem, Israel.
Gould, R. A., Otto, M. W. and Pollack (1995). A metaanalysis of treatment outcome for panic disorder.
Clinical Psychology Review, 15, 819±844.
Hamilton, M. (1959). The assessment of anxiety states by
rating. British Journal of Medical Psychology, 32, 50±55.
Hamilton, M. (1960). A rating scale for depression.
Journal of Neurology, Neurosurgery and Psychiatry, 23,
56±62.
Herbert, J. D. and Mueser, K. (1992). Eye movement
desensitization: A critique of the evidence. Journal
of Behaviour Therapy and Experimental Psychiatry, 23,
169±174.
Hertzberg, M. A., Feldman, M. E., Beckham, J. C. and
Davidson, R. T. (1996). Trial of trazadone for posttraumatic stress disorder using a multiple baseline
group design. Journal of Clinical Psychopharmacology, 16,
294±298.
Hickling, E. J. and Blanchard, E. B. (1997). The private
psychologist and manual-based treatments: Post-traumatic stress disorder secondary to motor vehicle
accidents. Behaviour Research and Therapy, 35, 191±203.
Horowitz, M., Wilner, N. and Alvarez, W. (1979). Impact
of Event Scale: A measure of subjective stress.
Psychological Medicine, 41, 209±218.
Hunter, J. and Schmidt, F. (1990). Methods of Meta
Analysis. Newbury Park, CA: Sage.
Jensen, J. A. (1994). An investigation of eye movement
desensitization and reprocessing (EMD/R) as a treatment for post-traumatic stress disorder (PTSD) symptoms of Vietnam combat veterans. Behaviour Therapy,
25, 311±325.
Kauffman, C. D., Reist, C., Djenderedjian, A., Nelson, J. N.
and Haier, R. J. (1987). Biological markers of affective
disorders and post-traumatic stress disorder: A pilot
Clin. Psychol. Psychother. 5, 126±144 (1998)
Meta-Analysis of PTSD Treatments
study with desipramine. Journal of Clinical Psychiatry,
48, 366±367.
Keane, T. M., Fairbank, J. A., Caddell, J. M. and Zimering,
R. T. (1989). Implosive (flooding) therapy reduces
symptoms of PTSD in Vietnam combat veterans.
Behaviour Therapy, 20, 245±260.
Keane, T. M., Caddell, J. M. and Taylor, K. L. (1988).
Mississippi scale for combat-related post-traumatic
stress disorder: Three studies in reliability and validity.
Journal of Consulting and Clinical Psychology, 56, 85±90.
Keane, T. M., Zimering, R. T. and Caddell, J. M. (1985). A
behavioural formulation of post-traumatic stress disorder. The Behaviour Therapist, 8, 9±12.
Kosten, T. R., Frank, J. B., Dan, E., McDougle, C. J. and
Giller, E. L. (1991). Drug therapy for post-traumatic
stress disorder using phenelzine or imipramine. Journal
of Nervous and Mental Disease, 179, 366±370.
Lambert, M., Hatch, D., Kingston, M. and Edwards, B.
(1986). Zung, Beck, and Hamilton rating scales as
measures of treatment outcome: A meta-analytic
comparison. Journal of Consulting and Clinical Psychology, 54, 54±59.
Lazrove, S., Triffleman, E., Kite, L., McGlashan, T. and
Rounsaville, B. (1996). The use of EMDR as treatment
for chronic PTSD: Encouraging results of an open trial.
Unpublished manuscript, Department of Psychiatry,
Yale University.
Lerer, B., Bleich, A., Kotler, M., Garb, R., Hertzberg, M.
and Levin, B. (1987). Post-traumatic stress disorder in
Israeli combat veterans. Archives of General Psychiatry,
44, 976±981.
Lilienfeld, S. (1996). EMDR treatment: Less than meets the
eye? The Skeptical Inquirer, 20, 25±31.
Lipper, S. (1990). Carbamazepine in the treatment of posttraumatic stress disorder: Implications for the kindling
hypothesis. In: M. E. Wolf and A. D. Mosnaim (Eds),
Post-traumatic Stress Disorder: Etiology, Phenomenology,
and Treatment. Washington, DC: American Psychiatric
Press, pp. 184±203.
Lipsey, M. and Wilson, D. (1993). The efficacy of
psychological, educational, and behavioural treatment:
Confirmation from meta-analysis. American Psychologist, 48, 1181±1209.
Litz, B. T. (1992). Emotional numbing in combat-related
post-traumatic stress disorder: A critical review and
reformulation. Clinical Psychology Review, 12, 417±432.
Litz, B. T. and Keane, T. M. (1989). Information processing
in anxiety disorders: Application to the understanding
of post-traumatic stress disorder. Clinical Psychology
Review, 9, 243±257.
Lohr, J., Kleinknecht, R. A., Tolin, D. F. and Barrett, R. H.
(1995). The empirical status of the clinical application of
eye movement desensitization and reprocessing. Journal
of Behaviour Therapy and Experimental Psychiatry, 26,
285±302.
Marcus, S., Marquis, P. and Sakai, C. (1996). Eye movement desensitization and reprocessing: A clinical outcome study for post-traumatic stress disorder.
Unpublished manuscript, Kaiser Adult Psychiatry,
Santa Clara, CA.
Marmar, C. R., Schoenfeld, F., Weiss, D. S., Metzler, T.,
Zatzick, D., Wu, R., Smiga, S., Tecott, L. and Neylan,
# 1998 John Wiley & Sons, Ltd.
143
T. (1996). Open trial of fluvoxamine treatment for
combat-related post-traumatic stress disorder. Journal of
Clinical Psychiatry, 57 (Suppl 8), 66±72.
McNally, R. J. (1996). Review of `Eye movement desensitization and reprocessing: Basic principles, protocols,
and procedures' by F. Shapiro. Anxiety, 2, 153±155.
Mollica, R. F., Wyshak, G., Lavelle, J., Truong, T., Tor, S.
and Yang, T. (1990). Assessing symptom change in
Southeast Asian refugee survivors of mass violence and
torture. American Journal of Psychiatry, 147, 83±88.
Montgomery, R. W. and Ayllon, T. (1994). Eye movement desensitization across subjects: Subjective and
physiological measures of treatment efficacy. Journal
of Behaviour Therapy and Experimental Psychiatry, 25,
217±230.
Mowrer, O. H. (1960). Learning Theory and Behaviour.
New York: Wiley.
Nagy, L. M., Morgan, C. A., Southwick, S. M. and
Charney, D. S. (1993). Open prospective trial of
fluoxetine for post-traumatic stress disorder. Journal of
Clinical Psychopharmacology, 13, 107±113.
Norris, F. H. (1992). Epidemiology of trauma: frequency
and impact of different potentially traumatic events on
different demographic groups. Journal of Consulting and
Clinical Psychology, 60, 409±419.
Orwin, R. (1983). A fail-safe N for effect size. Journal of
Educational Statistics, 8, 147±159.
Otto, M. W., Penava, S. J., Pollack, S. J. and Smoller,
J. W. (1996). Cognitive-behavioural and pharmacologic
perspectives on the treatment of post-traumatic
stress disorder. In: M. H. Pollack, M. W. Otto and
J. F. Rosenbaum (Eds), Challenges in Clinical Practice:
Pharmacologic and Psychosocial Strategies. New York:
Guilford, pp. 219±260.
Pitman, R. K., Orr, S. P., Altman, B., Longpre, R. E.,
Poire, R. E., Macklin, M. L., Michaels, M. J. and
Steketee, G. S. (1996a). Emotional processing and
outcome of imaginal flooding therapy in Vietnam
veterans with chronic post-traumatic stress disorder.
Comprehensive Psychiatry, 37, 409±418.
Pitman, R. K., Orr, S. P., Altman, B., Longpre, R. E., Poire,
R. E. and Macklin, M. L. (1996b). Emotional processing
during eye-movement desensitization and reprocessing
therapy of Vietnam veterans with chronic posttraumatic stress disorder. Comprehensive Psychiatry, 37,
419±429.
Reist, C., Kauffman, C. D., Haier, R. J., Sangdahl, C.,
DeMet, E. M., Chicz-DeMet, A. and Nelson, J. N.
(1989). A controlled trial of desipramine in 18 men with
post-traumatic stress disorder. American Journal of
Psychiatry, 146, 513±516.
Renfrey, G. and Spates, C.R. (1994). Eye movement
desensitization: A partial dismantling study. Journal
of Behaviour Therapy and Experimental Psychiatry, 25,
231±239.
Rescorla, R. A. (1988). Pavlovian conditioning: It's not
what you think it is. American Psychologist, 43,
151±160.
Richards, D. A., Lovell, K. and Marks, I. M. (1994).
Post-traumatic stress disorder: Evaluation of a
behavioural treatment program. Journal of Traumatic
Stress, 7, 669±680.
Clin. Psychol. Psychother. 5, 126±144 (1998)
144
Rosenthal, R. (1979). The `file drawer' problem and
tolerance for null results. Psychological Bulletin, 86,
638±641.
Rothbaum, B. O. (in press) A controlled study of EMDR
for PTSD. Bulletin of the Menninger Clinic.
Rothbaum, B. O., Foa, E. B., Riggs, D., Murdock, T. and
Walsh, W. (1992). A prospective examination of posttraumatic stress disorder in rape victims. Journal of
Traumatic Stress, 5, 455±475.
Rothbaum, B. O., Ninan, P. T. and Thomas, L. (1996).
Sertraline in the treatment of rape victims with posttraumatic stress disorder. Journal of Traumatic Stress, 9,
865±871.
Saunders, B. E., Arata, C. M. and Kilpatrick, D. G.
(1990). Development of a crime-related post-traumatic
stress disorder scale for women within the Symptom
Checklist-90-Revised. Journal of Traumatic Stress, 3,
439±448.
Scheck, M. M., Schaffer, J. A. and Gillette, C. S. (in press).
Brief psychological intervention with young high-risk
females: A comparison of eye-movement desensitization and reprocessing with active reflective listening.
Journal of Traumatic Stress.
Shapiro, F. (1989). Eye movement desensitization: A new
treatment for post-traumatic stress disorder. Journal
of Behavior Therapy and Experimental Psychiatry, 20,
211±217.
Shapiro, F. (1991). Eye movement desensitization and
reprocessing procedure: From EMD to EMD/R:A,
new treatment model for anxiety and related trauma.
The Behavior Therapist, 5, 128±133.
Shaprio, F. (1995). Eye Movement Desensitization and
Reprocessing: Basic Principles, Protocols, and Procedures.
New York: Guilford.
Shestatzky, M., Greenberg, D. and Lerer, B. (1988).
A controlled trial of phenelzine in post-traumatic stress
disorder. Psychiatry Research, 24, 149±155.
Smith, M. L., Glass, G. V. and Miller, T. I. (1980).
The Benefits of Psychotherapy. Baltimore, MD: Johns
Hopkins University Press.
Spielberger, C. D., Gorsuch, R. L. and Lushene, R. E.
(1970). Manual for the State-Trait Inventory. Palo Alto,
CA: Consulting Psychologists Press.
# 1998 John Wiley & Sons, Ltd.
M. L. Van Etten and S. Taylor
Sutherland, S. M. and Davidson, J. R. T. (1994). Drug
therapy for post-traumatic stress disorders. Psychiatric
Clinics of North America, 17, 409±423.
Taylor, S. (1995). Assessment of obsessions and
compulsions: Reliability, validity, and sensitivity
to treatment effects. Clinical Psychology Review, 15,
261±296.
Taylor, S. (1996). Meta-analysis of cognitive-behavioural
treatments for social phobia. Journal of Behaviour
Therapy and Experimental Psychiatry, 27, 1±9.
van der Kolk, B. (1987). The drug treatment of posttraumatic stress disorder. Journal of Affective Disorders,
13, 203±213.
van der Kolk, B. A., Dreyfuss, D., Michaels, M., Shera, D.,
Berkowitz, R., Fisler, R. and Saxe, G. (1994). Fluoxetine
in post-traumatic stress disorder. Journal of Clinical
Psychiatry, 55, 517±522.
Vaughan, K., Armstrong, M. S., Gold, R., O' Connor, N.,
Jenneke, W. and Tarrier, N. (1994). A trial of eye
movement desensitization compared to image habituation training and applied muscle relaxation in posttraumatic stress disorder. Journal of Behaviour Therapy
and Experimental Psychiatry, 25, 283±291.
Vaughan, K. and Tarrier, N. (1992). The use of image
habituation training with post-traumatic stress disorders. British Journal of Psychiatry, 161, 658±664.
Veronen, L. J., Kilpatrick, D. G. and Resick, P. A.
(1978). Stress inoculation training for victims of rape.
Paper presented at the annual meeting of the Association for Advancement of Behaviour Therapy, Chicago,
Illinois.
Wilson, S. A., Becker, L. A. and Tinker, R. H. (1995). Eye
movement desensitization and reprocessing (EMDR)
treatment for psychologically traumatized individuals.
Journal of Consulting and Clinical Psychology, 63,
928±937.
Wilson, S. A., Becker, L. A. and Tinker, R. H. (1997).
15-Month follow-up of EMDR treatment for psychological trauma. Journal of Consulting and Clinical
Psychology, 65, 1047±1056.
Wolf, F. M. (1986). Meta-analysis. Newbury Park, CA:
Sage.
Clin. Psychol. Psychother. 5, 126±144 (1998)